Phone Case OCR Code Reading
Optical character text recognition on textured injection-molded plastic
The ability to trace individual parts is increasingly important in modern supply chains. A simple way of ensuring an injection-molded object such as a phone case is correct is to emboss human-readable date and lot information at the point of manufacture. The code can then be used for inventory, counterfeit prevention, and identifying the source if the product later malfunctions. Unreadable embossed text impairs tracing capability, leading to a lack of transparency in the supply chain.
The embossed text is direct part marked (DPM), made of the same material as the case body. It is low contrast. The case interior is textured. Molding variations mean the text can have a range of reflectivities and shades. If there are even small defects in the injection molding, the result can be missing bits or distortions of the characters. Slight variations in lighting angle can lead to different appearances of the letters.
All of these issues lead to inaccuracies when this text is read by conventional machine vision OCR technology.
Cognex Deep Learning accurately reads challenging embossed OCR codes on fast-moving production lines. Deep Learning’s OCR tool is easily set up and deployed thanks to a pretrained font library. It trains on a small, labeled image set of embossed text on phone cases. Once the OCR tool has learned each character, it identifies each against the confusing case background, even with unpredictable molding defects, and with various lighting angles.
Whenever there is a new case material or text location, it takes only an updated labeled image set to train the OCR tool to identify and read the new text.